The CHAID tree displayed above is for a small sample run of 918 ZIP codes.
- The Base rates (the Dependent Variable) for each ZIP code are from a major carrier and are assumed to be based on credible experience. The Base rate and ZIP codes were segmented into 3 categories:
- Low ($130 to $150)
- Intermediate ($151 to $270)
- High ($271 to $450)
Note that our version of the CHAID model can work with up to 31 categories.
- Certain U.S. Census data for each ZIP code were selected as the Predictor Variables.
- Population per Square Mile: 17 categories topping out at 19,403.
- Percent of Population living in rural area: 6 categories.
- Average Vehicles per Square Mile: 20 categories topping out at 8,334.
- Average Number of Vehicles per Housing Unit: 6 categories topping out at 2.3.
- Average Number of Vehicles per Capita: 7 categories topping out at 1.1.
- Average Commute Travel Time: 9 categories topping out at 62.8 minutes.
Each U.S. Census statistic, here too, could have been segmented into 31 categories.
The group that proved to be most statistically significant in predicting "high BI base rates" consisted of the following features:
- Population per Square Mile: 4,620 and greater
- Average Vehicles per Capita: 0.6 or less
- Vehicles per Household unit: 1.4 and greater
All of the ZIP codes, in the statistically significant segmented group, had "high base rates"; none had intermediate or low base rates.
A group that proved to be most statistically significant in predicting "low BI base rates" consisted of the following features:
- Population per Square Mile: 89 to 1,241
- Percent of the Population living in Rural area: 75% and greater
None of the ZIP codes, in the statistically significant segmented group, had a "high base rate".
|
|
